An Optimization Ensemble Model for the Detection of Fake News

Abstract

The rising problem of misrepresentation and misinformation has taken center stage in the contemporary world as a consequence of the digitization of information. The rapid circulation of false information aggravates social and civic conflicts, erodes social trust, and stalls responsible public policy and governance. Therefore, distinguishing fraudulent information from real information has never been more critical. This research proposes a new integrated approach to the problem of fake news in considering the limitations posed by the machine learning approach. This research proposes an enhanced integrated model by making use of the distinctive advantages of several classifiers. For this purpose, the Firefly Algorithm is adopted to assign the optimal feature set and optimally distribute weights to the three base models: XGBoost, SVM, and LR integrated in the optimization ensemble model. The experimental results showed significant optimization ensemble performance improvement with an accuracy of 99.97%, a 99.96% F1-score, 99.94% precision, and 99.97% recall. The performance was comparable to, or slightly better than, that of the top individual model, XGBoost, demonstrating increased ensemble strength and robustness. These results endorse the approach of ensemble learning combined with ensemble learning